coordinator-init-mode
The complete two-phase Init Mode protocol the Squad coordinator runs when no team exists yet in the current repo. Phase 1 = propose the team (no files created, wait for user confirm). Phase 2 = create .squad/ scaffolding, casting state, .gitattributes for merge drivers, and the always-on built-ins (Scribe, Ralph, Rai, Fact Checker). Loaded on demand when the coordinator detects no .squad/team.md exists.
- License 许可
- MIT
- 所属仓库
- microsoft/waza
- 最近更新
- 2026-11-15
- Stars
- ★ 1.5k
- Forks
- 101
01 前言 Front Matter
- name
- coordinator-init-mode
- domain
- squad-internals
- source
- Extracted from squad.agent.md (bradygaster/squad#1308), with Waza project overrides for paired casting initialization and governance-aligned reviewer seeding.
- confidence
- high
- description
- The complete two-phase Init Mode protocol the Squad coordinator runs when no team exists yet in the current repo. Phase 1 = propose the team (no files created, wait for user confirm). Phase 2 = create .squad/ scaffolding, casting state, .gitattributes for merge drivers, and the always-on built-ins (Scribe, Ralph, Rai, Fact Checker). Loaded on demand when the coordinator detects no .squad/team.md exists.
- allowedTools
- []
03 SKILL.md 正文
Load this skill when: you detect that no
.squad/team.mdexists in the resolved team root — that means this is a fresh repo or a repo that has never been squadified, and Init Mode is the right path. The short stub insquad.agent.mdtells you to load this skill; the full two-phase protocol lives here.⚠️ Eager-execution exception: Init Mode is the ONE exception to the eager-execution / parallel-fan-out doctrine. Phase 1 MUST end with a user confirmation before any file is created. Do not bypass.
Phase 1: Propose the Team
No team exists yet. Propose one — but DO NOT create any files until the user confirms.
- Identify the user. Run
git config user.nameto learn who you're working with. Use their name in conversation (e.g., "Hey {user}, what are you building?"). Store their name (NOT email) inteam.mdunder Project Context. Never read or storegit config user.email— email addresses are PII and must not be written to committed files. - Ask: "What are you building? (language, stack, what it does)"
- Cast the team. Before proposing names, run the Casting & Persistent Naming algorithm (see the canonical Casting reference at
.squad/templates/casting-reference.md):- Determine team size: pick 4–5 cast (user-domain) agents, then add the 4 always-on built-ins (Scribe + Ralph + Rai + Fact Checker — see their dedicated sections in
squad.agent.md). A typical fresh squad has 8–9 total roster entries, not 4–5. - Determine assignment shape from the user's project description.
- Derive resonance signals from the session and repo context.
- Select a universe. Allocate character names from that universe.
- Scribe is always "Scribe" — exempt from casting.
- Ralph is always "Ralph" — exempt from casting.
- Rai is always "Rai" — exempt from casting.
- Fact Checker is always "Fact Checker" — exempt from casting (same pattern as Scribe / Ralph / Rai).
- Determine team size: pick 4–5 cast (user-domain) agents, then add the 4 always-on built-ins (Scribe + Ralph + Rai + Fact Checker — see their dedicated sections in
- Propose the team with their cast names. Example (names will vary per cast):
?️ {CastName1} — Lead Scope, decisions, code review
⚛️ {CastName2} — Frontend Dev React, UI, components
? {CastName3} — Backend Dev APIs, database, services
? {CastName4} — Tester Tests, quality, edge cases
? Scribe — (silent) Memory, decisions, session logs
? Ralph — (monitor) Work queue, backlog, keep-alive
?️ Rai — (background) RAI awareness, content safety
? Fact Checker — (verifier) Verification + Devil's Advocate
-
Use the
ask_usertool to confirm the roster. Provide choices so the user sees a selectable menu.The
questionmust be self-contained because the client may render the blocking input request without the assistant text that preceded it.The
questionmust reuse every roster line from the exact proposal emitted in step 4, in the same order. Do not regenerate, summarize, truncate, omit, or replace any roster row with an ellipsis. Every proposed member must appear with their emoji, cast name, role, and concise responsibility/scope — including all four always-on built-ins (Scribe, Ralph, Rai, Fact Checker).Shape:
Confirm this team (universe: {ActualUniverse}): {Every roster line from step 4, copied in full and in order} Look right?Never use a bare "Look right?" question and never rely on preceding assistant output to identify the team being confirmed.
- choices:
["Yes, hire this team", "Add someone", "Change a role"]
- choices:
⚠️ STOP. Your response ENDS here. Do NOT proceed to Phase 2. Do NOT create any files or directories. Wait for the user's reply.
Phase 2: Create the Team
Trigger: The user replied to Phase 1 with confirmation ("yes", "looks good", or similar affirmative), OR the user's reply to Phase 1 is a task (treat as implicit "yes").
If the user said "add someone" or "change a role," go back to Phase 1 step 3 and re-propose. Do NOT enter Phase 2 until the user confirms.
- Create the
.squad/directory structure (see.squad/templates/for format guides or use the standard structure:team.md,routing.md,ceremonies.md,decisions.md,decisions/inbox/,casting/,agents/,orchestration-log/,skills/,log/,rai/).
Casting state initialization — Waza project repair: Begin with a genuinely empty new target. Use .squad/casting-transactions.md's actual locked SDK writer to initialize schema-v1 registry/history and their commit manifest together, including complete provenance and the first assignment snapshot; only then initialize policy config exclusively from approved defaults. Do not create either casting JSON file independently or copy legacy examples into live state. On non-local backends, require an equivalent atomic operation or stop; generic per-file state writes are not sufficient.
Initialization requires a genuinely empty new target. The SDK prepares/validates existing files before the callback, so it cannot normalize v0.10 files left in place. Known frozen v0.10 state requires the guide's approved, quiesced archive-and-initialize procedure with disclosed downtime and continuity limits; never infer legacy status from a managed-pair validation failure.
Seeding: Create initial history.md context only where the agent's charter permits that output; Rai and Fact Checker do not create histories. Agent folder names are the cast name in lowercase (e.g., .squad/agents/ripley/). The Scribe's charter includes maintaining decisions.md and cross-agent context sharing. .squad/agents/rai/charter.md is seeded from Rai-charter.md, and .squad/rai/policy.md is seeded from rai-policy.md. .squad/agents/fact-checker/charter.md is seeded from fact-checker-charter.md and .squad/fact-checker/policy.md is seeded from fact-checker-policy.md.
Waza reviewer seeding override (Squad only): Use the project-repaired templates
under .squad/templates/, not pristine upstream replacements. Seeded review policy
must agree with .squad/governance.md and the support charters: rejection includes
concrete counter-evidence, a different qualified revision owner, and focused
rereview rather than author/rejecting-gate pairing. Ordinary Copilot work does
not enter Squad Init Mode or acquire these coordination rules.
After successful focused rereview of the independently corrected revision, the selected
gate may issue approval/PASS, not native repository approval, merge or publishing authority.
Charter output/persistence restrictions take precedence over generic history, audit-trail or
decision-inbox instructions. Return findings in the current review response; only an authorized
coordinator or Scribe handles required persistence under the current state contract.
Team.md structure: team.md MUST contain a section titled exactly ## Members (not "## Team Roster" or other variations) containing the roster table. This header is hard-coded in GitHub workflows (squad-heartbeat.yml, squad-issue-assign.yml, squad-triage.yml, sync-squad-labels.yml) for label automation. If the header is missing or titled differently, label routing breaks.
Merge driver for append-only files: Create or update .gitattributes at the repo root to enable conflict-free merging of .squad/ state across branches:
.squad/decisions.md merge=union
.squad/agents/*/history.md merge=union
.squad/log/** merge=union
.squad/orchestration-log/** merge=union
.squad/rai/audit-trail.md merge=union
The union merge driver keeps all lines from both sides, which is correct for append-only files. This makes worktree-local strategy work seamlessly when branches merge — decisions, memories, and logs from all branches combine automatically.
-
Say: "✅ Team hired. Try: '{FirstCastName}, set up the project structure'"
-
Post-setup input sources (optional — ask after team is created, not during casting):
- PRD/spec: "Do you have a PRD or spec document? (file path, paste it, or skip)" → If provided, follow PRD Mode flow.
- GitHub issues: "Is there a GitHub repo with issues I should pull from? (owner/repo, or skip)" → If provided, follow GitHub Issues Mode flow.
- Waza membership policy: Propose AI members only under
.squad/governance.md; do not add human members or human issue/implementation handoffs. Preserve the project owner/requester and required repository approvals. - Copilot agent: "Want to include @copilot? It can pick up issues autonomously. (yes/no)" → If yes, follow Copilot Coding Agent Member section and ask about auto-assignment.
- These are additive. Don't block — if the user skips or gives a task instead, proceed immediately.
04 同仓库更多 Skills
waza-runner
Run evaluations on Agent Skills to measure their effectiveness. USE FOR: "run skill evals", "evaluate my skill", "test skill quality", "check skill triggers", "skill compliance check", "measure skill performance", "run evals on [skill-name]", "grade skill execution". DO NOT USE FOR: writing skills (use skill-authoring), improving frontmatter (use sensei), or general testing unrelated to skills.
waza
**WORKFLOW SKILL** - Evaluate AI agent skills using structured benchmarks with YAML specs, fixture isolation, and pluggable validators. USE FOR: run waza, waza help, run eval, run benchmark, evaluate skill, test agent, generate eval suite, init eval, compare results, score agent, agent evaluation, skill testing, cross-model comparison. DO NOT USE FOR: improving skill frontmatter (use waza dev), creating new skills from scratch (use skill-creator), token counting or budget checks (use waza tokens). INVOKES: Copilot SDK executor, mock engine, code/regex validators. FOR SINGLE OPERATIONS: use waza run directly for a single benchmark.
waza-interactive
Interactive workflow partner for creating, testing, and improving AI agent skills with waza. USE FOR: run my evals, check my skill, compare models, create eval suite, debug failing tests, is my skill ready, ship readiness, interpret results, improve score. DO NOT USE FOR: general coding, non-skill work, writing skill content (use skill-authoring), improving frontmatter only (use sensei).
code-explainer
Explains code snippets in plain English, breaking down what the code does step by step. Perfect for learning, code reviews, or documentation.
compatibility-skill
Read fixture status and report whether it is ready.
evidence-judge
Read a supplied report and summarize its recorded status and numeric items.